Free 2D to 3D photo converter · runs entirely in your browser
From flat photo to stereo pair
Drop an image, tune the three parameters and pick an output format.
Drag an image here
or — JPG, PNG, WebP
Drag multiple images here
or — JPG, PNG, WebP
Got an image series or animation frames in a folder?
Every image inside — including subfolders — gets added in order.
3D Parameters
Depth Anything V2 — Small downloads ~25MB once and is cached by your browser.
Applies to every image in the batch. Change it, then click Process all again to re-render images you've already processed.
How much displacement is applied between near and far.
The screen plane: what "pops out" and what "sinks in".
Virtual distance between the two stereo cameras.
Blurs the raw model output before warping — softens artifacts, softens the effect.
Grid sampling is fast and hole-free — good while adjusting sliders. Mesh warping warps a triangle mesh built from the depth map and can look more three-dimensional, at a slower recompute.
Suggestion
Pick an image to get a tailored tip.
How it works
A four-stage pipeline
Follows SBS-2DTo3D / ComfyUI-Y7-SBS-2Dto3D's depth → depth-image-based-rendering approach, with the model and every rendering step running client-side via ONNX Runtime Web.
-
01
Neural depth estimation
The image is run through Depth Anything V2, a real monocular depth model, executed in the browser via Transformers.js/ONNX Runtime Web on WebGPU (or WASM as a fallback) — the same model SBS-2DTo3D uses, no server round-trip.
-
02
Depth map refinement
The model's output is optionally smoothed and can be inverted, to soften noise or flip which side "pops."
-
03
Stereo view synthesis
The left/right pair is rendered straight from the depth map using one of two depth-image-based-rendering methods: grid sampling, which backward-samples each output pixel so there are no holes to fill, or mesh warping, which displaces a triangle mesh built over the image so neighbouring cells stretch to cover disocclusions — no separate inpainting pass needed either way.
-
04
Stereo composition
The two views are merged into the chosen format: parallel side-by-side, cross-eye, or red-cyan anaglyph.
Formats
An output for every way of looking
Parallel SBS
Two images side by side, left eye on the left. The standard format for most VR headsets.
Cross-eye SBS
Right eye on the left: made for viewers who cross their gaze to see it with the naked eye, no headset needed.
Red-Cyan Anaglyph
A single dual-channel image, viewed with classic red-cyan glasses.